Article Preview
Top1. Introduction
Predictive analysis is the procedure of gathering information from data sets to discover future results. The pregnancy delivery mode prediction is significant for physicians to categorize the fundamental factors such as cesarean births or normal birth. There is various invention related to predict the types of delivery during pregnancy. However, medical researchers report that the better prediction of the type of childbirth still faces many challenges. Though cesarean and normal delivery is used generally, there are diverse risks and difficulties related to cesarean delivery. Besides, prediction is used for prior preparations of financial support and pain control. Thus, the machine learning methods are employed for better identification of delivery types by considering medical parameters.
(Ana Raquel Neves, 2017) identify various factors associated with the mode of delivery. (AlessioPetrozziello, 2019) implemented Multimodal Convolutional Neural Network (MCNN) model to analyze the big available database for predicting the cesarean forceps or ventouse delivery using patient records. (MorganeLinard, 2019) introduced Robson classification method to forecast the cesarean section or vaginal delivery using logistic regression models. However, the conventional failed to analyze more clinical risk factors for delivery prediction. This leads to provide minimal accuracy in the pregnancy delivery mode prediction.
Therefore, Dichotomous Radial Basis Tanimoto Network Prediction (DRBTNP) Method is proposed to enhance the process of pregnancy delivery mode prediction with higher accuracy. The proposed DRBTNP method comprises different types of layers for performing delivery mode prediction with less time and space utilization. At first, the number of features of patient data is considered as input. Input layer transmits image to the hidden layer for measuring similarity between two data. In DRBTNP method, tanimoto similarity computation is carried out to lessen the prediction time of delivery modes. During this process, similarity between training and testing data is estimated. Then the resultant similarity score is given to the output layer where the prediction is carried out. In this layer, normal and cesarean delivery is predicted by analyzing the similarity score from hidden layer. This is done by applying activation function called radial basis function in the output layer. In radial basis function, the comparison between similarity score with threshold value is performed to predict the delivery mode. The similarity score result is greater than threshold, patient data is predicted as cesarean. When similarity score is minimal than threshold, the patient data is predicted as normal or vaginal. In this way, the women pregnancy delivery mode is effectively predicted with higher accuracy.
Dichotomous Radial Basis Tanimoto Network Prediction (DRBTNP) Method is proposed for pregnancy delivery mode prediction with maximum accuracy and less time. The proposed DRBTNP method contains the three different layers. At first, patient data with selected features is utilized as input. Then these inputs are forwarded into hidden layer. In the hidden layer, tanimoto similarity function is used identify the relationship between training and testing data. The output of tanimoto similarity function is varied from 0 to 1 where 0 represents the low similarity and 1 represents the higher similarity. The output of similarity score is send to output layer. An output layer employs activation function to identify cesarean and normal delivery. In DRBTNP technique, radial basis function is utilized to predict the given data as normal delivery or cesarean delivery. The results from hidden layer (similarity score) is evaluated with threshold value in radial basis function. When similarity score is greater than threshold, then the data is predicted as cesarean. Else, similarity score is minimal, and then data is predicted as normal. Therefore, pregnancy delivery mode prediction is effectively achieved in proposed DRBTNP technique.
DRBTNP method is implemented in Java Language. When conducting the experiments, proposed techniques utilize the patient data from National Center for Health Statistics (NCHS). During the experimentation, the different number of patient data is used and it varied as 100 to 1000 for ten iterations. As shown in experimental results, proposed techniques effectively perform pregnancy delivery mode prediction with enhanced results in prediction accuracy, prediction time and space utilization. Proposed DRBTNP method improves the of prediction accuracy by 9% when compared to existing methods. The proposed DRBTNP method minimizes the prediction time 20% and space utilization by 19% when compared to conventional works.